Climate-Driven Shifts in Drought Dynamics in the Balkhash–Alakol Water Management Basin Revealed by Integrated Satellite and Ground Observations

Abstract

Droughts represent a major threat to water security and agricultural sustainability in the arid regions of Central Asia. The Balkhash–Alakol water management basin is particularly vulnerable due to its complex runoff formation, dependence on surface water resources, and the transboundary nature of the Ile River, the main inflow to Lake Balkhash. This study analyzes hydrometeorological data for 1950–2023 using the Standardized Precipitation Evapotranspiration Index (SPEI), Streamflow Drought Index (SDI), and Surface Water Supply Index (SWSI). Satellite-based indicators, including the Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST), were used to derive the Vegetation Condition Index (VCI), Temperature Condition Index (TCI), and Vegetation Health Index (VHI) for 2002–2023. Results indicate a significant increase in drought frequency and duration since the 1990s, with SPEI values reaching −2.6 during extreme events. Satellite observations revealed widespread vegetation degradation, with drought-affected areas (VHI < 30) exceeding 60% of the basin in severe years. Strong correlations between meteorological and hydrological drought indices (r up to 0.9) highlight the importance of cumulative moisture deficits. The results indicate an increase in the region’s aridity and the manifestation of drought in climatic, hydrological, and ecosystem changes. The practical significance of this study lies in the potential to use these results to improve drought monitoring and early warning systems, assess water risks, and develop measures for adapting to climate change and making decisions regarding water resource management in the Balkash–Alakol water management basin.

1. Introduction

Drought is among the most complex natural phenomena, significantly impacting natural ecosystem functioning, water resources, and the socio-economic development of various global regions [1,2]. Unlike most hydrometeorological extremes, drought develops gradually and is characterized by a complex formative nature driven by the interaction of climatic, hydrological, and biophysical factors [3,4]. Its slow onset and multifactorial development processes significantly hinder timely identification and quantitative assessment of drought conditions. Under global climate change, the risk of drought conditions is intensifying, primarily due to rising air temperatures and increased evapotranspiration. These factors lead to heightened moisture deficits and an increase in the frequency, duration, and intensity of dry periods across many regions of the world [5,6]. The most pronounced increases in drought severity have been reported in the Mediterranean region, southwestern North America, southern Africa, southwestern Australia, and Central Asia, where climate change has exacerbated water scarcity, ecosystem degradation, and agricultural vulnerability [5,7,8,9].

Central Asia is a region of high climatic vulnerability due to its continental climate, significant spatiotemporal variability of precipitation, and the heavy reliance of water resources on snow-and-glacier melt from mountain systems [10,11]. The water regime of most rivers in the region is primarily formed by snow and glacier melt in mountainous areas, making water resources particularly sensitive to changes in air temperature and precipitation patterns [12,13]. Under conditions of high climatic variability and increasing anthropogenic pressure on water management systems, drought phenomena acquire not only a natural but also a pronounced socio-economic character, significantly impacting agricultural sustainability, ecosystem health, and the functioning of regional water infrastructure [14,15]. In recent years, this has been confirmed by several studies analyzing the spatiotemporal dynamics of drought conditions in Central Asia. Specifically, Aitekeyeva et al. [16], using the Vegetation Health Index (VHI) derived from MODIS data for Central Asian agricultural lands, demonstrated that droughts regularly encompass vast areas of the region’s agro-landscapes. According to the study, in certain years, the proportion of agricultural land affected by drought conditions exceeded approximately 40% of the total area, confirming the high vulnerability of agrarian ecosystems to climatic variability. Similar results have been obtained in other studies, noting an increase in the frequency and intensity of drought episodes, as well as the intensification of vegetation and ecosystem degradation processes in the arid regions of Central Asia [17,18].

Among the countries of Central Asia, the Republic of Kazakhstan holds a distinct position regarding the impact of climate variability on water resources. A significant portion of the country’s territory is characterized by arid and semi-arid climatic conditions, while the availability of surface water resources largely depends on transboundary river basins [19,20]. The Balkhash–Alakol water management basin holds particular significance within the water management system of southeastern Kazakhstan, with the transboundary Ile River playing a key role by forming the bulk of the inflow into Lake Balkhash [21,22]. The formation of the basin’s water resources is closely linked to the snow-and-glacier melt of the Tian Shan Mountain regions located within China and Kazakhstan, resulting in high sensitivity of the river hydrological regime to changes in air temperature and precipitation patterns [9,18].

Rising air temperatures in mountainous regions accelerate glacier degradation and alter snowmelt processes, affecting river runoff generation and the seasonal distribution of water resources. Concurrently, in lowland areas, temperature increases are accompanied by rising evapotranspiration and intensifying moisture deficits, creating preconditions for the development of drought conditions. This results in an amplified hydro-climatic contrast between mountainous headwater catchments, where the bulk of river runoff is generated, and lowland regions, where the primary water consumption is concentrated. Under climate change, such processes may lead to an increase in the frequency and intensity of droughts, as well as shifts in the spatiotemporal structure of the region’s water resources [23,24]. Changes in hydro-climatic conditions and river runoff regimes significantly impact the state of deltaic ecosystems, vegetation productivity, and the sustainability of wetlands [25,26]. Over the last two decades, signs of wetland ecosystem degradation, declining vegetation productivity, and intensifying desertification processes have been observed within the basin. These changes indicate the increasing vulnerability of the region’s natural ecosystems and necessitate a detailed analysis of the spatiotemporal patterns of drought development [18].

Assessment and monitoring of drought conditions are essential tools for analyzing hydro-climatic variability and managing water resources under climate change. To quantitatively evaluate drought conditions, a wide range of indices has been developed, enabling the analysis of various drought types, including meteorological, hydrological, and agricultural. One of the most widely used indicators is the Standardized Precipitation Index (SPI), based on the analysis of atmospheric precipitation and the identification of moisture anomalies across different time scales [27,28]. To account for the impact of air temperature and evaporation processes, the Standardized Precipitation Evapotranspiration Index (SPEI) was proposed, which incorporates potential evapotranspiration and more accurately reflects water balance changes under global warming [29]. For the analysis of hydrological droughts, the Streamflow Drought Index (SDI), based on river runoff data [30], and the Surface Water Supply Index (SWSI), which considers the cumulative impact of precipitation, snow cover, river runoff, and reservoirs on water resource formation, are employed [31].

Alongside traditional climatic and hydrological drought indices, Earth remote sensing methods have gained widespread application in recent decades, enabling the assessment of vegetation status and surface conditions across extensive territories. Satellite data provide spatially continuous observations, which is particularly vital for regions with limited meteorological station networks. One of the most prevalent indicators is the Normalized Difference Vegetation Index (NDVI), extensively used to evaluate vegetation condition and identify stress conditions related to moisture deficits [32]. Based on NDVI and surface temperature characteristics, derivative indices such as Vegetation Condition Index (VCI), Temperature Condition Index (TCI), and Vegetation Health Index (VHI) were developed to assess the degree of moisture and temperature stress in vegetation [33]. The use of satellite-derived indices allows for the identification of spatial patterns in drought development and the analysis of ecosystem responses to changes in hydro-climatic conditions.

In recent years, a number of studies have been conducted to evaluate the impact of climate change on water resources and aridity in the region under consideration. Specifically, it has been established that meteorological and hydrological droughts significantly influence the water level dynamics of Lake Balkhash and may lead to alterations in the basin’s hydrological regime under climate change conditions [22]. Furthermore, analysis of the climatic indices SPI and SPEI revealed trends in the intensity and frequency of drought phenomena in the southern Balkhash region [24]. Despite previous studies addressing individual aspects of drought and water resources in the Balkhash–Alakol Basin, a comprehensive assessment of the spatio-temporal dynamics of drought conditions based on the integrated use of climatic, hydrological, and satellite-derived drought indices has not yet been conducted. Unlike previous studies, which primarily focused on individual drought types or specific parts of the basin, the present study combines meteorological, hydrological, and remote sensing data for the entire Balkhash–Alakol water management basin. This integrated approach provides a more comprehensive assessment of the spatio-temporal characteristics of drought development and their impacts on water resources and ecosystem conditions under changing climate conditions.

Thus, the objective of this study is to evaluate the spatiotemporal dynamics of drought conditions in the Balkhash–Alakol water management basin through a comprehensive analysis of hydro-climatic and satellite drought indices. To achieve this, the spatiotemporal dynamics of drought conditions were analyzed using climatic (SPEI), hydrological (SDI, SWSI), and satellite (NDVI, VCI, VHI) indices within the Balkhash–Alakol water management basin. The results enable the identification of spatiotemporal patterns in drought development and an assessment of their impact on water resource formation and ecosystem status within the basin. Integrated analysis of climatic, hydrological, and satellite drought indices facilitates more accurate identification of dry periods and their spatial distribution, which is essential for monitoring hydro-climatic risks. The study findings may be utilized to improve drought monitoring systems, enhance water resource management efficiency, and develop adaptation measures under climate change.

2. Materials and Methods

2.1. Study Area

The catchment area of the Balkhash–Alakol water management basin covers approximately 413,000 km², of which about 15% is located within the northwestern part of the Xinjiang Uygur Autonomous Region of China [21]. Within the basin, there are over 52,000 rivers and ephemeral streams; approximately 90% of these belong to the Lake Balkhash basin, while the remainder pertains to the Alakol lake system [21]. The primary rivers flowing into Lake Balkhash include the Ile, Karatal, Aksu, Lepsy, and Ayagoz [21] (Figure 1).

The Ile River is the primary water artery of the basin and the main source of water inflow into Lake Balkhash. The total length of the river is 1439 km, with 815 km flowing through Kazakhstan, and the catchment area reaching approximately 131,000 km² [21,22,34]. The major part of runoff generation occurs in China, where the hydrographic network is characterized by higher density compared to the middle and lower reaches of the basin. The river’s runoff formation is primarily associated with snow-and-glacier melt from the Tian Shan Mountain regions, making the basin’s water resources sensitive to changes in air temperature and precipitation patterns [34].

The Ile River is predominantly fed by snow and glacier melt, with most of its runoff generated by the melting of seasonal snow cover and glaciers in the Tien Shan Mountains. The long-term mean river discharge is approximately 450 m³/s. The highest mean monthly discharge occurs in July and reaches about 700 m³/s, whereas the lowest mean monthly discharge is observed in January and is approximately 250 m³/s [35]. A characteristic feature of the river’s hydrological regime is a pronounced summer high-flow period caused by intensive snow and glacier melt. At present, the river discharge regime is also influenced by climate change and flow regulation by the Kapchagay Reservoir [21,34].

From a hydrological perspective, the basin area includes a runoff generation zone in the mountain regions of the Tian Shan and Zhetysu Alatau, as well as a zone of runoff distribution and loss in the lowland part of the basin. In high-altitude areas (above 3000 m), runoff generation is predominantly determined by glacier-and-snow melt, while in mid-altitude areas, seasonal snowmelt and atmospheric precipitation play a significant role. In the lowland part of the basin, runoff is formed by snowmelt, atmospheric precipitation, and groundwater recharge [21,22,34].

The Balkhash–Alakol water management basin is one of the key regions of Kazakhstan in terms of water resources; however, it is simultaneously among the territories most sensitive to climate change. In recent decades, substantial changes in hydrometeorological conditions have been observed within the basin, accompanied by a transformation of the intra-annual distribution of river discharge. Talipova et al. [34] demonstrated that the current hydrological regime of the Ile River basin is shaped by the combined effects of climate change and anthropogenic factors, including flow regulation and increasing water withdrawals. Alimkulov et al. [22] reported that the water balance of Lake Balkhash is highly sensitive to changes in river inflow and evaporation, while scenario-based modeling indicates a potential reduction in inflow and a consequent decline in lake water levels. Abdrakhimov et al. [36] also identified changes in the intra-annual distribution of river discharge, reflecting a transformation of the seasonal flow regime. These findings indicate that increasing air temperature, changes in precipitation patterns, intensified evaporation, and anthropogenic flow regulation are the main drivers of increasing drought severity and the transformation of the hydrological regime of the basin’s rivers.

2.2. Research Materials

This study used ground-based meteorological observations, hydrological measurements, and satellite remote sensing data. All meteorological and hydrological drought indices were calculated using a common reference period (1950–2023). Long-term observations from 16 meteorological stations located within the Balkhash–Alakol water management basin, including air temperature and precipitation data, were used to assess meteorological drought conditions. Hydrological drought was evaluated using streamflow records from four hydrological stations located on the main rivers of the basin, allowing the analysis of long-term variability in water resources and the identification of hydrological drought characteristics (Figure 2).

Satellite products eMODIS and eVIIRS, obtained from the USGS EarthExplorer platform, were used for the spatial analysis of drought conditions [37,38,39]. The analysis included 16-day composite eMODIS NDVI V6 and eMODIS Global LST V6 datasets for the period 2002–2022, based on Collection 6 source data, as well as the corresponding 10-day composite eVIIRS NDVI and eVIIRS Global LST products for 2023, generated following the discontinuation of eMODIS production on 1 October 2022 [37,38,39]. The eMODIS datasets were derived from MODIS observations acquired by the Terra and Aqua satellites, whereas the eVIIRS datasets were derived from VIIRS observations acquired by the Suomi NPP and NOAA-20 satellites [37,38,39]. All datasets were provided in GeoTIFF format with a spatial resolution of approximately 1 km and were referenced to a common coordinate system.

Since the eMODIS and eVIIRS products are distributed as standard preprocessed datasets, atmospheric correction, Quality Control (QC), cloud screening, and other standard preprocessing procedures had already been applied according to the official product specifications prior to data distribution. Atmospheric correction for both product streams was performed upstream by the data provider (USGS EROS), following standard MODIS/VIIRS surface-reflectance and land-surface-temperature algorithms [37]. Quality control and cloud screening were implemented through the accompanying per-pixel quality and acquisition-date layers distributed with each composite [40], which flag cloud-, cloud-shadow-, and snow/ice-contaminated observations; only pixels flagged as usable (“good”/“nominal” quality) were retained during the respective 16-day (eMODIS) or 10-day (eVIIRS) compositing procedure, which selects the least cloud-affected, highest-quality observation within each product’s compositing period. These quality flags and acquisition-date layers were used in the present study to exclude compromised pixels prior to the calculation of the VCI, TCI, and VHI indices. To ensure spatial consistency, the eVIIRS products were resampled to match the eMODIS grid (1 km) using bilinear interpolation.

The continuity of the long-term NDVI, LST, and VHI time series was evaluated by examining the transition period (2022–2023). A comparison of basin-averaged VHI values, the drought-affected area (VHI < 30), and the spatial distribution of drought conditions before and after the transition from eMODIS to eVIIRS did not indicate any apparent discontinuities or systematic shifts in the time series. These findings indicate good consistency between the two satellite products and support the suitability of the merged dataset for long-term drought monitoring in the Balkhash–Alakol Basin.

2.3. Research Methods

The drought assessment in the Balkhash–Alakolwater management basin was conducted based on an integrated analysis of climatic, hydrological, and satellite drought indices. This approach allows for the consideration of various aspects of aridity formation, including meteorological, hydrological, and ecological processes (Figure 3).

The drought classification criteria presented in Figure 3 were adopted from the original methodologies and widely accepted classification schemes reported in the scientific literature. The classification thresholds for the SPEI, SDI, and SWSI indices were adopted from the approaches proposed by [29,30,31], respectively. The classification of the satellite-derived indices VCI, TCI, and VHI is based on the methodology developed by [33,41], which has been widely applied for drought monitoring and vegetation condition assessment using remote sensing data. The NDVI threshold values were adopted in accordance with the recommendations of [32,42], which describe the assessment of vegetation condition and the identification of drought conditions based on satellite observations.

Standardized Precipitation Evapotranspiration Index (SPEI). Among the recent approaches designed to identify climatic droughts, the SPEI stands out. This index is based on the calculation of daily precipitation and air temperature time series (maximum and minimum air temperature). The procedure for determining the index value fully follows the calculation procedure for the SPI, but in addition to precipitation, surface temperature is also taken into account.

The proposed SPEI is calculated using a procedure similar to that of the SPI. However, instead of precipitation, the calculation of SPEI in formula (2) utilizes the differences (D) between monthly precipitation totals (R) and potential evapotranspiration (PET):

$$D_i=R_i-{PET}_i$$

(1)

where i is the ordinal number of the calculated month.

The SPEI was calculated using the Climpact web application following the methodology of Vicente-Serrano et al. [29]. The index is based on the climatic water balance (P − PET), where potential evapotranspiration (PET) was estimated from maximum and minimum air temperature data, taking geographical latitude into account [43]. The climatic water balance was standardized using a three-parameter log-logistic probability distribution, while the entire available observation period (1950–2023) was used as the reference period.

Streamflow Drought Index (SDI). The SDI, developed by Nalbantis and Tsakiris [30], is widely used to assess hydrological drought at different time scales. In this study, the SDI was calculated from monthly river discharge observations obtained at hydrological gauging stations. Using monthly river discharge data (V_km), the SDI can be calculated for different reference periods within the hydrological year using the following equation:

$${\mathrm{S}\mathrm{D}\mathrm{I}}_{i, k}={\displaystyle \frac{V_{i, k }- V_{km}}{S_k}}$$

(2)

where i = 1, 2; …, and k = 1, 2, 3, 4. V_km and S_k are the mean and standard deviation of the cumulative river runoff volumes for the baseline period k. k = 1 for October–December, k = 2 for October–March, k = 3 for October–June and k = 4 for October–September. In this study, the hydrological year was defined from October to September in accordance with the original methodology proposed by Nalbantis and Tsakiris [30], ensuring that the complete annual runoff cycle was represented within a single hydrological year.

Surface Water Supply Index (SWSI). To assess surface water availability, the SWSI was used as an integrated indicator that accounts for the combined influence of the main components of the basin water balance [31]. In its original formulation, the SWSI can be calculated using standardized values of river discharge, precipitation, snow water equivalent (SWE), and reservoir storage, depending on the hydrological characteristics of the basin and the availability of input data.

In the present study, the SWSI was calculated using monthly river discharge and precipitation data. Snow water equivalent (SWE) and Kapshagay Reservoir storage were not included in the calculations because continuous and homogeneous observation records were unavailable for the entire study period (1950–2023). Nevertheless, the effects of snowmelt and reservoir regulation are indirectly represented through the observed river discharge, which integrates the dominant hydrological processes controlling water resource formation within the basin.

Although both the SDI and SWSI are hydrological drought indices, they describe different aspects of hydrological variability. The SDI is based exclusively on river discharge data and is designed to identify streamflow anomalies associated with the development of hydrological drought in river systems. In contrast, the SWSI characterizes the overall availability of surface water resources within a basin by integrating river discharge with other components of the water balance, the composition of which depends on the hydrological characteristics of the basin and the availability of input data. The combined use of the SDI and SWSI makes it possible to distinguish between streamflow deficits and the overall status of surface water availability, thereby providing a more comprehensive assessment of hydrological drought conditions in the Balkhash–Alakol Basin.

Normalized Difference Vegetation Index (NDVI). The NDVI was used to characterize vegetation conditions. This index reflects the ratio between radiation absorption in the red band and reflection in the near-infrared band of the spectrum, enabling the assessment of changes in vegetation density and productivity under moisture-deficit conditions [32,33]. Land Surface Temperature (LST) was used to characterize the thermal conditions of the land surface and the energy balance of the land–atmosphere system.

The raw satellite NDVI and LST data were converted to physical values using standard scaling factors:

$$NDVI=DN\times 0.0001$$

(3)

$$LST\left(°\mathrm{C}\right)=DN\times 0.02-273.15$$

(4)

where DN is the digital value of the pixel.

The scaling coefficients used to convert the Digital Number (DN) values into physical values of the NDVI and LST were adopted according to the official specifications of the corresponding satellite products. For NDVI, the standard scaling factor of 0.0001 was applied to convert the digital values into dimensionless index values, consistent with the official NASA MODIS Vegetation Index product specifications [44]. For LST, the standard scaling factor of 0.02 was used, followed by conversion from Kelvin to degrees Celsius by subtracting 273.15, consistent with the USGS EROS eMODIS/eVIIRS product documentation [37]. The use of these scaling coefficients ensures accurate derivation of the physical variables required for subsequent calculation of the VCI, TCI, and VHI indices [33,41].

After conversion to physical values, only NDVI values representing actual vegetation conditions were retained for further analysis. To minimize the influence of noise and non-vegetated surfaces, only NDVI values within the range of 0.05–0.85 were included. Pixels with NDVI values < 0.05, corresponding to water bodies, bare soils, or noise effects, and NDVI values > 0.85, representing potential anomalies, were excluded from further calculations.

Consequently, the VHI analysis primarily represents areas with detectable vegetation cover rather than completely barren surfaces. Nevertheless, naturally sparse vegetation in arid and semi-arid environments may also exhibit persistently low VHI values. Therefore, low VHI values were interpreted as indicators of vegetation stress rather than direct evidence of ecosystem degradation or chronic drought. Land-cover stratification was beyond the scope of the present study but represents an important direction for future research to further improve the interpretation of VHI patterns.

LST was converted from digital values to degrees Celsius following the standard procedure described above, after which a filter of permissible temperature values corresponding to the actual climatic conditions of the study area was applied. For LST, a physically permissible surface temperature range of −40…+60 °C was used; values outside this range were considered anomalous and excluded from the analysis.

Calculation of Drought Indices (VCI, TCI and VHI). To quantitatively evaluate the degree of moisture and temperature stress on the vegetation cover in the Balkhash–Alakolwater management basin, the methodology proposed by A. Kogan [33] was used. This methodology is based on the normalization of vegetation and thermal parameters relative to their long-term extreme values.

The Vegetation Condition Index (VCI), which characterizes the degree of vegetation stress, was calculated using the formula:

$$\mathrm{V}\mathrm{C}\mathrm{I}={\displaystyle \frac{{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{i}}-\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}\mathrm{m}\mathrm{i}\mathrm{n}}{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}\mathrm{m}\mathrm{a}\mathrm{x} -\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}\mathrm{m}\mathrm{i}\mathrm{n}}} \times 100$$

(5)

where NDVI_i is the NDVI value at the current time (year), and NDVI_min and NDVI_max are the long-term minimum and maximum NDVI values determined for each pixel over the corresponding observation period.

The Temperature Condition Index (TCI), reflecting the degree of temperature stress on vegetation, was determined as follows:

$$\mathrm{T}\mathrm{C}\mathrm{I}={\displaystyle \frac{\mathrm{L}\mathrm{S}\mathrm{T}\mathrm{m}\mathrm{a}\mathrm{x} -{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{i}}}{\mathrm{L}\mathrm{S}\mathrm{T}\mathrm{m}\mathrm{a}\mathrm{x}-\mathrm{L}\mathrm{S}\mathrm{T}\mathrm{m}\mathrm{i}\mathrm{n}}} \times 100$$

(6)

where LST_i is the land surface temperature value at the current time, and LST_min and LST_max are the long-term minimum and maximum LST values for each pixel.

The integral Vegetation Health Index (VHI), which combines the effects of moisture and temperature stress, was calculated as a weighted combination of the VCI and TCI indices with equal weighting coefficients (0.5 and 0.5), following the original formulation of Kogan [33,41]:

$${VHI}_i=0.5\times {VCI}_i+0.5\times {TCI}_i$$

(7)

In this study, the VHI was calculated using the original formulation proposed by Kogan [33,41], in which the VCI and the TCI are assigned equal weighting coefficients (0.5 and 0.5). This formulation represents the standard VHI methodology and has been extensively applied in satellite-based drought monitoring and vegetation health assessment studies.

Although several studies have explored the possibility of adjusting the weighting coefficients according to regional climatic and environmental conditions, no universally accepted methodology currently exists for determining region-specific weights for arid environments such as those of Central Asia. Moreover, the application of alternative weighting schemes requires comprehensive regional calibration and independent validation, which would reduce the comparability of results across different studies.

Therefore, the original VHI formulation with equal weighting coefficients for the VCI and TCI components was adopted in this study. This approach ensures methodological consistency, facilitates direct comparison with previous international studies, and follows the internationally accepted standard methodology for VHI-based drought assessment [33,41].

To ensure the consistency of the long-term satellite time series, all datasets were harmonized to a common spatial resolution (1 km), map projection, and coordinate reference system prior to analysis. The spatial resolution of the eVIIRS data was resampled to match that of the eMODIS data (1 km) using the bilinear interpolation method. Furthermore, all satellite datasets underwent identical preprocessing procedures, including quality control and the exclusion of invalid observations, to ensure temporal consistency between the two satellite products. As a result, a unified VHI time series covering the 2002–2023 period was generated and used to analyze the interannual dynamics of drought conditions.

The VHI threshold values (<10, 10–20, 20–30, 30–40, and >40) were adopted according to the classical classification proposed by Kogan [33,41], which has been widely applied in satellite-based drought monitoring studies. This classification categorizes vegetation drought severity into five levels, ranging from extreme drought to the absence of drought conditions. The threshold values were not calibrated specifically for the study area but were adopted to maintain methodological consistency and ensure the comparability of the results with previous VHI-based drought assessments [33,41].

The area of drought-affected territories for each year was calculated as the ratio of the total area of pixels satisfying the condition VHI < 30 to the total valid area of the basin:

$${\mathrm{D}\mathrm{r}\mathrm{o}\mathrm{u}\mathrm{g}\mathrm{h}\mathrm{t}\mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}}_{\mathrm{i}}={\displaystyle \frac{\mathrm{A}\left(\mathrm{V}\mathrm{H}\mathrm{I}\mathrm{i}<30\right)}{\mathrm{A}\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{d}}} \times 100\%$$

(8)

where A(VHI_i < 30) is the total area of pixels with VHI < 30 in the i-th year, and A_valid is the total valid area of the basin.

To evaluate the spatial stability and recurrence of drought conditions, the drought frequency was calculated for each pixel as the number of years during which a VHI value < 30 was recorded:

$$F={\displaystyle {\sum }_{i=1}^N}I\left({VHI}_i<30\right)$$

(9)

where F is the drought frequency for a given pixel; N is the total number of observation years; VHI_i is the VHI value in the i-th year; and I is the indicator function, which takes the value of 1 if the condition is met (VHI_i < 30) and 0 otherwise.

Based on the resulting drought frequency map, the basin territory was classified by climate risk levels (low, moderate, high, and chronic aridity), which allowed the identification of zones with sustained recurrence of drought conditions and persistent vegetation stress. Additionally, long-term average VHI values were calculated to identify the background vulnerability of the territory and spatial patterns of climate stress distribution.

All calculations for area and drought frequency were performed in ArcGIS (ArcMap) 10.8 (Esri Inc., Redlands, CA, USA) using spatial analysis tools.

Long-Term Spatial Change Analysis of VHI.

To identify long-term spatial changes in vegetation condition, a spatial change analysis of the VHI was performed by comparing two representative periods: the baseline period (2002–2010) and the recent period (2016–2023).

Average long-term VHI values were calculated for each period. Subsequently, a difference map of changes was constructed using the formula:

$$∆\mathrm{V}\mathrm{H}\mathrm{I}={\overline{\mathrm{V}\mathrm{H}\mathrm{I}}}_{2016-2023}-{\overline{\mathrm{V}\mathrm{H}\mathrm{I}}}_{2002-2010}$$

(10)

The threshold of ΔVHI = ±5 was adopted as an empirical classification criterion to distinguish areas with noticeable improvement or deterioration in vegetation condition between the two study periods. This threshold was used for descriptive classification purposes and does not represent a statistical significance threshold.

The resulting difference raster reflects the direction and magnitude of changes in the state of the vegetation cover.

To interpret the results, a threshold classification of changes was applied:

ΔVHI ≤ −5—pronounced deterioration of vegetation status;

−5 <ΔVHI <5—relative stability;

ΔVHI ≥ 5—pronounced improvement.

To exclude insignificant interannual fluctuations, only spatially significant changes in vegetation status, where the absolute change in the index exceeded the ΔVHI ≥ 5 threshold, were further identified. This approach minimizes the influence of random interannual variability and highlights stable spatial trends in aridity changes. Based on the resulting classification, a map of spatial changes in the state of the vegetation cover was constructed.

Additionally, the areas and proportions of the territory for each change class were calculated by summing the number of pixels in the corresponding category and normalizing them to the total valid area of the basin.

Statistical Methods for Trend Analysis. To identify statistically significant trends and evaluate their temporal structure in hydrometeorological data series, the Mann–Kendall test was used, representing a non-parametric method widely applied for the analysis of monotonic trends in time series [45,46]. To determine potential regime shift points, a sequential Mann–Kendall test based on the analysis of the forward sequential statistic (UF) and the backward sequential statistic (UB) was applied, allowing the identification of structural changes in the time series [47].

To quantitatively evaluate the trend magnitude, Sen’s slope estimator was applied, which allows for determining the rate of change in the parameter over time and remains robust against outliers and non-normality of data distribution [48].

Pettitt’s test was used to identify structural change points in drought index time series. Pettitt’s test is a nonparametric rank-based method developed by Pettitt [49] for detecting a single change point (structural break) in a time series. The test identifies statistically significant shifts in the median of a series, which may indicate changes in the drought regime associated with climatic variability or anthropogenic influences. A major advantage of the method is that it does not require the data to follow a specific probability distribution and is highly sensitive to changes in the central tendency, making it particularly suitable for hydrological time series, which often violate the assumption of homogeneity. The test is based on rank statistics and evaluates the null hypothesis of no change point against the alternative hypothesis that a significant shift in the median has occurred at an unknown point in the time series [49].

3. Results

3.1. Long-Term Changes in Key Meteorological Parameters

The results of the Mann–Kendall test application revealed a statistically significant positive air temperature trend at all studied meteorological stations for the 1950–2023 period. The rise in air temperature varies from 0.19 °C/10 years (MS Ulken Almaty and Sarkand) to 0.43 °C/10 years (MS Bakanas); in all cases, the trend is significant (p < 0.001) (

Table 1. Results of trend analysis (Mann–Kendall test and Sen’s slope estimator) for air temperature and atmospheric precipitation for the 1950–2023 period.

Meteorological Stations	Air Temperature			Atmospheric Precipitation
	Test Z	p-Value	Sen’s Slope	Test Z	p-Value	Sen’s Slope
Narynkol	6.94	***	0.029	−0.63	ns	−0.261
Kyrgyzsay	5.28	***	0.022	1.25	ns	0.714
Almaty	7.40	***	0.042	0.45	ns	0.279
Esik	6.08	***	0.029	−1.08	ns	−0.829
Ulken Almaty	5.49	***	0.019	0.05	ns	0.044
Aidarly	5.84	***	0.035	−0.92	ns	−0.340
Kuygan	6.66	***	0.041	0.00	ns	0.002
Aul-4	6.20	***	0.038	−2.46	*	−0.480
Shelek	7.28	***	0.038	0.38	ns	0.138
Zharkent	6.99	***	0.036	1.05	ns	0.298
Bakanas	6.85	***	0.043	−0.54	ns	−0.210
Lepsi	6.91	***	0.035	−0.68	ns	−0.480
Sarkand	3.73	***	0.019	−0.74	ns	−0.358
Saryozek	5.64	***	0.025	1.39	ns	0.581
Kogaly	5.73	***	0.023	0.24	ns	0.198
Ushtobe	6.39	***	0.038	0.87	ns	0.296

*** p < 0.001; * p < 0.05; ns—not significant.

).

In contrast to air temperature, atmospheric precipitation dynamics are characterized by pronounced spatial heterogeneity and are statistically insignificant in most cases (p > 0.05) Z-statistic values vary from −1.08 to 1.39, indicating the absence of a single directed trend in precipitation change. The exception is the Aul-4 meteorological station, for which a statistically significant negative trend in atmospheric precipitation was identified (Z = −2.46, p < 0.05).

Next, to evaluate long-term dynamics and identify potential change points (inflections) in the air temperature and atmospheric precipitation series for the 1950–2023 period, a sequential Mann–Kendall test was applied. This method, based on time-series analysis, allows for the identification of monotonic trends and the determination of possible shifts in trend characteristics. Analysis of the sequential UF and UB statistics showed the presence of a sustained positive air temperature trend at all studied meteorological stations.

During the early stages of the time series, the UF statistic values do not exceed the thresholds and fluctuate near zero, indicating the absence of a significant trend. In the modern period, a sustained increase in UF is observed, exceeding the critical boundaries (±1.96; p <0.05, which points to a statistically significant rise in air temperature. The intersection of the UF and UB curves is interpreted as a potential change point in trend characteristics. For most meteorological stations, such intersections are observed in the late 1980s and early 1990s. For the Aidarly, Bakanas, Ushtobe, Zharkent, and Kuygan meteorological stations, the intersection of the UF and UB curves indicates a potential change point occurring between 1989 and 1991, suggesting a synchronous modification of the temperature regime across the region. For the high-altitude stations Esik and Ulken Almaty, the potential change point is identified slightly later, in the mid-1990s, which can be attributed to local physiographic factors, including altitudinal zonation and orographic effects (Figure 4).

Following the identified change points, a sustained increase in the UF statistic values is observed, indicating an intensification of the positive air temperature trend in the subsequent period. The results obtained are consistent with regional and global assessments of modern climate warming, which has been particularly pronounced since the late 20th century. For a more detailed analysis, a comparative assessment of the temperature regime was performed for two time intervals (before and after the 1990s), identified based on the sequential Mann–Kendall test results.

It has been established that in the modern period, the highest intensity of positive temperature trends is observed during the spring-summer season, where values reach up to 1.6 °C/10 years (MS Bakanas). In contrast to the warm season, the cold period of the year is characterized by multi-directional temperature change dynamics. In particular, during the autumn-winter months, negative trend values are recorded at several meteorological stations. For instance, at MS Esik in November, values decrease by −0.8 °C/10 years, and at MS Saryozek, by up to −0.6 °C/10 years.

Thus, the intra-annual distribution of temperature trends is characterized by a pronounced seasonal asymmetry during the modern period (post-1990), manifesting as an intensification of positive tendencies in the spring-summer period and their weakening, down to negative values, in the cold season. Such a structure of changes indicates a redistribution of the thermal regime throughout the year and reflects the transformation of the region’s climatic system (Figure 5).

Time-series analysis of atmospheric precipitation using the sequential Mann–Kendall test allowed for the evaluation of its long-term variability and trend direction, as well as the determination of potential change points. Comparison of the sequential UF and UB statistics showed that for most meteorological stations, a sustained statistically significant trend in annual precipitation totals is absent (Figure 6).

At the majority of meteorological stations, time intervals are identified within which a decrease in annual atmospheric precipitation totals is observed. In particular, following the established regime shift points (1970s–early 1990s), a transition of the UF statistic values into the negative range is noted, indicating a prevailing tendency toward a decrease in precipitation. However, in most cases, the UF values do not exceed the critical thresholds (±1.96), which indicates that the identified changes are statistically insignificant at the p < 0.05 significance level. At certain meteorological stations (Esik, Saryozek), the long-term dynamics of atmospheric precipitation are unstable, characterized by alternating positive and negative deviations.

In contrast to air temperature, which is characterized by a sustained positive trend during the modern period, the atmospheric precipitation regime within the studied basin is characterized by the absence of a pronounced statistically significant trend, despite the presence of individual periods of decline. Combined with the identified sustained rise in air temperature, this leads to an intensification of potential evapotranspiration and the formation of a moisture deficit, contributing to the development of aridization processes and an increased probability of drought conditions.

The analysis of the intra-annual distribution of precipitation revealed pronounced spatial and temporal variability, with substantial differences among individual months and meteorological stations. Before the regime shift, positive changes predominated at most stations, particularly during individual months of the spring and summer seasons. Following the regime shift, a redistribution of the intra-annual precipitation pattern was observed; however, no consistent decreasing trend was identified throughout the entire warm season. At the Bakanas station, precipitation decreased mainly in June, as well as during the spring (April–May) and autumn (September–October) months. At the Aidarly station, negative changes were observed in April, May, November, and December, while positive tendencies persisted in July and August. At the Esik station, the most pronounced decrease occurred in July, whereas at the Saryozek station precipitation decreased from May to August, with the greatest decline observed in August. These results indicate considerable spatial heterogeneity in precipitation changes and a redistribution of precipitation throughout the year, which, together with rising air temperatures, may contribute to the intensification of drought conditions during certain periods of the growing season (Figure 7).

3.2. Long-Term Dynamics of Drought Conditions

The results of the long-term dynamics analysis showed that the hydroclimatic regime of the studied region is characterized by high interannual variability, with alternating dry and wet phases of varying intensity. The amplitude of index fluctuations reaches values from −3.11 (MS Ushtobe), corresponding to extreme drought, to +2.65 (MS Bakanas), reflecting high moisture conditions, which indicates significant contrast in climatic conditions (Figure 8 and Figure 9).

Analysis of the SPEI-6 index, characterizing moisture availability conditions on a medium-term timescale, showed that a pronounced interannual variability with alternating dry and wet phases of varying intensity is observed at all studied meteorological stations. Throughout the period under consideration, recurring episodes of moderate (SPEI-6 < −1.0), severe (SPEI-6 < −1.5), and extreme drought (SPEI-6 < −2.0) were recorded at most stations. Furthermore, starting from the late 1980s and early 1990s, a sustained trend toward an increase in the frequency and duration of dry periods is traced, manifesting as a rise in the number of consecutive years with negative index values.

The longest drought periods are noted at the lowland meteorological stations. For instance, a sustained drought episode was recorded at MS Aul-4 in 2004–2008, lasting up to 5 years, during which SPEI-6 values corresponded to severe and, in some places, extreme drought. Periods of similar duration are observed at MS Bakanas, Lepsi, Ushtobe, and Kuygan, where an increase in the recurrence of drought conditions and a higher frequency of severe and extreme drought cases have been noted since the 1990s.

At MS Zharkent, Kogaly, and Shelek, prolonged drought episodes are also recorded, accompanied by index values below −1.5, indicating an intensifying moisture deficit. During certain years, SPEI-6 values reach −2.0 or lower, corresponding to extreme drought conditions, although their duration is generally shorter compared to the lowland areas.

Overall, the SPEI-6 analysis reveals pronounced spatial heterogeneity in drought conditions across the Balkhash–Alakol basin. For most meteorological stations, the period after the 1990s is characterized by an increased frequency of severe and extreme droughts; however, the magnitude of these changes varies considerably across the basin. The most pronounced intensification of drought conditions is observed in the central and western parts of the basin, whereas no substantial increase in drought duration is evident at the Kogaly, Zharkent, and Shelek stations. These spatial differences are primarily associated with the influence of the surrounding mountain systems, which enhance orographic precipitation, sustain snow- and glacier-fed runoff, and mitigate drought severity in the eastern and southeastern parts of the basin. In contrast, the central and western lowland areas are characterized by a more arid climate, higher atmospheric evaporative demand, and limited moisture availability, making them more susceptible to the development of severe and persistent drought conditions.

Analysis of the SPEI-12 index, reflecting the long-term dynamics of moisture availability, showed a predominance of sustained multi-year phases of aridity and moisture, forming under the influence of accumulated moisture deficit or surplus. Compared to the SPEI-6 index, this indicator is characterized by lower interannual variability and allows for a clearer identification of prolonged hydroclimatic anomalies.

At the majority of meteorological stations, a trend toward an increase in the duration of dry periods (SPEI-12 < −1.0) is observed, particularly over the last two decades. The duration of individual aridity phases reaches 4–6 years. The most sustained and prolonged dry periods are characteristic of lowland stations. In particular, at MS Bakanas and Aul-4, a continuous period of aridity is recorded for 2018–2023, accompanied by a predominance of negative index values and episodes of severe drought (SPEI-12 < −1.5). Similar patterns are observed at MS Kuygan, Ushtobe, and Lepsi, where the dominance of negative index values has been noted in recent years.

To further assess potential shifts in meteorological drought conditions, the Pettitt test was applied to the SPEI-12 index (

Table 2. Pettitt test results for the SPEI-12 index.

Meteorological Station	SPEI-12
Narynkol	2004 *
Kyrgyzsay	1974 ns
Almaty	1991 **
Esik	1993 *
Ulken Almaty	1991 *
Aidarly	1989 *
Kuygan	1989 ***
Aul-4	1988 ***
Shelek	2005 ns
Zharkent	1973 ns
Bakanas	1994 **
Lepsy	1973 *
Sarkand	2006 ns
Saryozek	1973 **
Kogaly	2004 *
Ushtobe	1994 **

Note: *** p < 0.001; ** p < 0.01; * p < 0.05; ns—not significant.

). The results indicate that statistically significant change points at most meteorological stations occurred during the late 1980s and early 1990s (1988–1994), generally coinciding with the period of intensified air temperature increase identified earlier. At several stations, statistically significant change points were either absent or occurred during earlier or later periods, reflecting spatial heterogeneity in the transformation of the meteorological drought regime.

The obtained results indicate a statistical correspondence between changes in the temperature regime and the long-term dynamics of the SPEI-12 index. Nevertheless, the coincidence of the detected change points should not be interpreted as evidence of a direct causal relationship, since drought development is governed by the combined effects of rising air temperature, precipitation variability, and evaporative demand.

Analysis of the SDI showed that the characteristics of hydrological drought manifestation differ significantly depending on the hydrological and physiographic conditions of the watersheds. As an example, four gauging stations characterizing various hydrological conditions of the studied region were examined (Figure 10).

Small mountain watercourses, such as the Kaskelen River (Kaskelen city) and the Kishi Almaty River (Almaty city), are characterized by high sensitivity to climatic factors, which manifests as significant interannual runoff variability and an increased recurrence of drought conditions. Since the 1980s, these rivers have exhibited more frequent episodes of hydrological drought (SDI < −1.0), with index values occasionally reaching levels corresponding to severe and extreme drought (SDI < −2.0). The duration of dry phases generally ranges from 2 to 6 years, indicating the formation of sustained runoff deficits.

In contrast, the Koktal River (Araltobe village) is characterized by more stable hydrological dynamics, where pronounced and prolonged periods of hydrological drought are virtually unobserved. The SDI values predominantly fluctuate near zero and positive values, indicating relatively stable water availability.

A unique type of hydrological regime is characteristic of the Ile River (37 km downstream of Kapshagay Reservoir), where anthropogenic runoff regulation associated with the operation of the Kapshagay Reservoir plays a significant role. During its filling period (the 1970s–early 1980s), a decline in water yield was observed due to runoff accumulation, which was accompanied by the formation of drought conditions of varying intensity (SDI < −1.0). In the subsequent period, the reservoir’s influence led to a smoothing of interannual runoff variability and a decrease in the amplitude of index fluctuations.

However, since the 2010s, an increase in the frequency of drought episodes has been noted for the Ile River, primarily due to climatic factors. Under modern conditions, despite the regulatory effect of the reservoir, SDI and SWSI values during certain periods decrease to levels corresponding to severe and extreme hydrological drought.

The Pettitt test results (

Table 3. Pettitt test results for hydrological drought indices (SDI and SWSI).

Hydrological Station	SDI	SWSI
Ile River–37 km downstream of Kapshagay Reservoir	1996 ns	1992 *
Kaskelen River–Kaskelen city	1980 **	1982 ***
Kishi Almaty River–Almaty city	1972 **	1973 ***
Koktal River-Araltobe village	1991 ***	1997 ***

Note: *** p < 0.001; ** p < 0.01; * p < 0.05; ns—not significant.

) indicate that statistically significant change points in the hydrological drought regime were identified only at a limited number of hydrological stations. For the SDI, significant change points were detected at the Kaskelen River–Kaskelen village (1980), Kishi Almaty River–Almaty city (1972), and Koktal River–Araltobe village (1991) stations, whereas no statistically significant change point was identified at the Ile River–37 km downstream of Kapshagay Reservoir. For the SWSI, statistically significant change points were detected at all analyzed hydrological stations, with the identified shifts occurring between 1973 and 1997.

The differences in the timing of the detected change points indicate that no common shift in the hydrological drought regime occurred across the study area. This finding highlights the spatial variability of hydrological regime changes, reflecting differences in runoff generation processes and the varying sensitivity of individual catchments to climatic and anthropogenic influences.

3.3. Integration of Ground-Based Observations and Earth Remote Sensing Data to Assess Drought Conditions

Given the identified patterns of long-term drought dynamics based on the analysis of ground-based observations, it is appropriate to supplement the study with modern Earth remote sensing methods. This allows for a more detailed examination of the spatial characteristics of drought processes and an evaluation of their development in recent decades.

Due to the limited availability of satellite data for earlier periods, the subsequent analysis was performed for the modern period (from 2002 onward), ensuring the comparability of results and a higher spatial resolution of the assessment.

Spatiotemporal drought dynamics according to the VHI. Analysis of the VHI for the 2002–2023 period revealed a pronounced interannual variability of drought conditions within the Balkhash–Alakolwater management basin. The proportion of drought-affected areas with moderate and more severe drought (VHI < 30) varied significantly by year, indicating an alternation of dry and relatively favorable climatic periods (Figure 11).

Remote sensing data shall be consistent with the results of ground-based observations: according to the SPEI-6 index in 2008, negative values ranging from −1.5 to −2.0 and below were recorded at most meteorological stations, corresponding to severe and, in some places, extreme drought. Specifically, according to data from MS Aul-4 (−2.11), MS Lepsi (−2.52), and MS Kuygan (−2.16), SPEI-6 values in 2008 reached the level of extreme drought, indicating a pronounced moisture deficit. This confirms that the drought conditions of 2008 were regional in nature and were manifested in both satellite and ground-based data.

Comparable scales of drought conditions were also observed in 2008 (69.7%), 2015 (49.2%), 2022 (60.2%), and 2023 (59.2%), indicating widespread vegetation and temperature stress. Visual analysis of VHI maps shows that in these years, the drought was not local but regional, affecting both lowland and foothill areas.

Moderate drought conditions (30–35% of the basin area) were noted in 2005, 2011, 2014, and 2019. A number of years were characterized by relatively favorable moisture conditions: in 2003, 2006, 2016, and 2017, the proportion of drought-affected territories remained minimal and did not exceed 15% of the basin area.

At the beginning of the study period (2002), drought conditions were primarily local in nature and were limited to certain areas in the central and northern parts of the basin. In subsequent years, an alternation of relatively favorable and extremely dry periods was observed.

However, in the final years of the study period, the nature of droughts changed significantly. The graph of the interannual dynamics of the drought-affected area (VHI < 30) shows an intensification and increased frequency of drought conditions starting around 2020. While during the 2002–2019 period, extreme droughts were predominantly episodic and alternated with years of vegetation recovery, a series of consecutive drought years has been observed after 2020.

In 2020, 2022, and 2023, the proportion of territory covered by moderate, severe, and extreme drought exceeded 47–60.2 of the valid basin area. Such values indicate not only individual extreme events but may also point to the formation of a more sustained regime of climatic stress.

Long-term average VHI value and baseline state of the territory. The long-term average value of the VHI for the 2002–2023 period was used to assess the general baseline state of the vegetation cover in the Balkhash–AlakolBasin. In contrast to the analysis of individual years, this indicator reflects the averaged level of vegetation and temperature stress across the basin throughout the entire study period, rather than short-term extreme events.

The results showed that most of the basin is characterized by moderate VHI values, indicating a predominance of chronic, yet not extreme, vegetation stress. The largest proportion of the territory corresponds to the moderate drought class (VHI = 20–30), which covers approximately 51.9% of the basin area. Furthermore, 37.3% of the territory falls under the severe drought class (VHI = 10–20), evidencing the widespread prevalence of unfavorable conditions for vegetation cover functioning.

The mild drought class (VHI = 30–40) occupies about 7.5% of the territory, while areas with extremely low VHI values (<10) have limited distribution, accounting for approximately 0.3%. Territories showing no signs of drought (VHI ≥ 40) are practically non-existent, occupying only 0.06% of the basin area.

It shall be noted that the long-term average VHI value reflects the integral state of ecosystems rather than the intensity of individual drought episodes. When averaged over a long-time interval, interannual fluctuations are smoothed: years with pronounced drought and years with more favorable conditions partially offset each other. Thus, during the driest years—2008, 2015, 2020, 2022, and 2023—VHI values decreased significantly across a large part of the basin, whereas in more favorable years, such as 2003, 2006, 2016, and 2017, a partial recovery of the vegetation cover state was observed.

As a result, the long-term average VHI indicator reflects a sustained baseline regime of ecosystem functioning rather than individual extreme years, smoothing out both drought peaks and recovery periods. This explains why, despite the occurrence of several large-scale droughts, the final long-term assessment is characterized predominantly by moderate index values.

Overall, the results obtained indicate that a sustained baseline of vegetation stress has formed within the basin during the study period, which shall be viewed as a manifestation of chronic climatic tension in the ecosystems of the arid and semi-arid zones.

Drought frequency and chronic risk zones. Analysis of drought year frequency, based on the summation of binary drought maps (VHI < 30) for the 2002–2023 period, allowed for the identification of zones with varying degrees of climatic risk (Figure 12,

Table 4. Drought frequency (VHI < 30) and territory area distribution for the 2002–2023 period.

Frequency (Years)	Area (km2)	Area (%)	Interpretation
0–3	111,429	25.3	Stable areas
4–7	140,374	32	Low drought frequency
8–11	119,622	27.2	Moderate risk
12–15	54,172	12.3	High risk
16–24	13,680	3.1	Chronic drought

). The results obtained show that only about 25.3% of the basin territory is characterized by stable conditions with rare drought occurrences (0–3 drought years during the observation period).

A significant portion of the basin (approximately 32%) was subject to recurring droughts with a low but regular frequency (4–7 years). Moderate-risk zones (8–11 years) cover 27.2% of the territory, while areas with a high frequency of droughts (12–15 years) account for 12.3% of the basin area. Chronic aridity zones, where drought conditions were observed for 16–24 years (more than 67% of the observation period), occupy a relatively small but significant share—about 3.1% of the territory.

The most vulnerable zones are concentrated in the central and northern parts of the basin, represented mainly by desert and semi-desert landscapes. These areas exhibit persistent vegetation stress and can be considered priority areas for drought monitoring under ongoing climatic stress.

Long-Term Spatial Change Analysis of VHI. Spatial analysis of long-term changes in the VHI between the baseline (2002–2010) and recent (2016–2023) periods revealed pronounced spatial heterogeneity (Figure 13). A deterioration in vegetation condition was observed across 32.1% of the basin area, while 43.6% remained relatively stable and 26.0% showed an improvement in VHI values.

In Figure 13b, only areas with pronounced negative changes (ΔVHI ≤ −5) are presented, allowing clearer identification of zones experiencing persistent vegetation stress. These areas represent spatially consistent patterns of reduced vegetation condition.

Trend analysis results. For the drought-affected area (VHI < 30): according to the results of the Mann–Kendall test for the proportion of the territory subject to drought (VHI < 30) for the 2002–2023 period, a statistically significant increasing trend was identified (τ = 0.34; Z = 2.10; p = 0.05) (

Table 5. Results of trend analysis for VHI and drought-affected area (VHI < 30).

Indicator	τ (Kendall)	Z	p-Value	Sen’s Slope
Mean VHI	−0.18	−1.28	>0.1	−0.25 VHI/year
Drought area (VHI < 30)	+0.34	2.10	0.05	+0.7%/year

). The Sen’s slope estimate shows that the drought-affected area increases on average by 0.7% of the basin area per year (Figure 14). This indicates a gradual expansion of the territories subject to drought.

For the average VHI value: according to the results of the Mann–Kendall test for the interannual series of the average VHI (Mean VHI) for 2002–2023, a weak downward trend was identified (τ = −0.18, Z = −1.28), which is not statistically significant (p > 0.1). The Sen’s slope estimate showed an average rate of decline of approximately −0.25 VHI units per year. This indicates a general but poorly expressed deterioration in the state of the vegetation cover, which is likely smoothed out by the high interannual variability of moisture and temperature conditions.

Furthermore, to assess the consistency of the drought index dynamics, a correlation analysis was performed. The results of the Pearson linear correlation analysis showed the presence of statistically significant positive relationships between all considered indices (p < 0.01). The highest correlation was established between the SDI and SWSI indices (r = 0.95), indicating a high level of consistency among indicators characterizing river discharge and the availability of surface water resources (

Table 6. Pearson correlation coefficients between drought indices (2002–2023).

	VHI	TCI	VCI	SDI	SWSI	SPEI-6	SPEI-12
VHI	1
TCI	0.84 **	1
VCI	0.65 **	0.17	1
SDI	0.63 **	0.36	0.69 **	1
SWSI	0.58 **	0.38	0.56 **	0.95 **	1
SPEI-6	0.67 **	0.35	0.78 **	0.84 **	0.72 **	1
SPEI-12	0.70 **	0.43 *	0.73 **	0.93 **	0.89 **	0.91 **	1

Note: ** p < 0.01; * p < 0.05.

).

Strong correlation links shall also be identified between SDI and SPEI-12 (r = 0.93), as well as between SWSI and SPEI-12 (r = 0.89), indicating the determining role of accumulated moisture deficit in the formation of hydrological drought. The high correlation between SPEI-6 and SPEI-12 (r = 0.91) reflects the consistency of drought processes across various time scales and their persistence.

Analysis of satellite indices showed that the VHI is closely linked to the TCI (r = 0.84), indicating the significant influence of temperature anomalies on the state of the vegetation cover. The relationship between VHI and the Vegetation Condition Index (VCI) is characterized by a moderately high degree (r = 0.65), reflecting the contribution of the water factor and vegetation productivity.

The relationship between the VHI and the SPEI meteorological indices is characterized by a moderately high degree (r = 0.67–0.70), while the correlation of VHI with hydrological indices is somewhat lower (r = 0.58–0.63). This evidences a more complex and indirect response of the vegetation cover to changes in moisture availability and river water yield.

Pearson correlation analysis revealed strong relationships among the investigated drought indices, particularly between the hydrological drought index (SDI) and the long-term meteorological drought index (SPEI-12) (Table 6). However, these correlation coefficients represent only the contemporaneous relationships between the indices and do not account for the temporal lag in the propagation of meteorological drought through the hydrological system. Since river discharge responds to atmospheric moisture deficits with a certain delay rather than instantaneously, an additional lagged correlation analysis between the SPEI and SDI indices was performed.

The lagged correlation analysis showed that, across all investigated river basins, the highest correlation coefficients between the SPEI and SDI indices occurred at positive time lags (

Table 7. Lagged Pearson correlation coefficients between the SPEI and SDI indices.

Hydrological Station	Lag 0	Lag 1	Lag 2	Lag 3	Lag 4	Lag 5	Lag 6	Lag 7	Lag 8	Lag 9	Lag 10	Lag 11	Lag 12
Ile River–37 km downstream of Kapshagay Reservoir	0.32	0.36	0.4	0.42	0.44	0.44	0.44	0.43	0.43	0.41	0.38	0.35	0.32
Kaskelen River–Kaskelen village	0.49	0.53	0.55	0.56	0.55	0.53	0.51	0.49	0.46	0.44	0.41	0.37	0.34
Kishi Almaty River–Almaty	0.41	0.45	0.47	0.48	0.47	0.46	0.45	0.43	0.4	0.37	0.33	0.29	0.25
Koktal River–Araltobe	0.62	0.68	0.71	0.71	0.7	0.68	0.65	0.61	0.56	0.51	0.45	0.39	0.33

Note: All correlation coefficients are statistically significant at p < 0.01.

), indicating that hydrological drought develops with a delay relative to meteorological drought. The maximum correlation coefficients ranged from 0.44 to 0.71.

For the Kaskelen, Kishi Almaty, and Koktal rivers, the strongest correlations were observed at a lag of 3 months, indicating a relatively rapid propagation of meteorological moisture deficits into streamflow drought. In contrast, for the Ile River downstream of the Kapshagay Reservoir, the maximum correlation coefficient (r = 0.44) was maintained over lags of 4–6 months, indicating a more prolonged hydrological response. This delayed response is primarily attributed to the large drainage area, the snow- and glacier-fed runoff regime, and the regulating effect of the Kapshagay Reservoir, which increases the storage capacity of the basin and prolongs the transmission of meteorological drought signals into river discharge.

4. Discussion

The results indicate a pronounced intensification of drought conditions in the Balkhash–Alakol Water Management Basin over recent decades. Since the late 1980s and early 1990s, most meteorological stations have recorded an increase in the frequency of severe and extreme drought events, consistent with the contemporary trends in hydroclimatic change observed across Central Asia. The obtained findings are consistent with previous studies conducted in the Ile River basin and the Lake Balkhash basin, which demonstrated that recent changes in the hydrological regime and water balance are driven by the combined effects of climate change, river flow regulation, and increasing water demand [22,23,24,34].

The intensification of drought conditions is primarily associated with changes in the basin water balance driven by rising air temperatures. Despite the absence of statistically significant long-term changes in precipitation, increasing temperatures enhance potential evapotranspiration, resulting in reduced effective moisture availability. Similar conclusions have been reported for Central Asia, where increasing atmospheric evaporative demand associated with warming has been identified as one of the primary drivers of drought development, even under relatively stable precipitation conditions [14,29,50].

The identified spatial patterns of drought development indicate a heterogeneous response of different parts of the Balkhash–Alakol Water Management Basin to contemporary climate change. The most pronounced intensification of drought conditions is observed in the central and western parts of the basin, whereas the eastern and southeastern regions exhibit less pronounced changes. This spatial differentiation is primarily controlled by the physiographic characteristics of the basin. The eastern part is influenced by the Zhetysu Alatau and Northern Tien Shan Mountain systems, where enhanced orographic precipitation, snow- and glacier-fed river systems, and lower air temperatures contribute to sustaining regional water resources. In contrast, the central and western parts of the basin are characterized by a more arid climate, higher atmospheric evaporative demand, and limited moisture supply, making them more vulnerable to the development of drought conditions [34,51,52].

A key characteristic of the Balkhash–Alakol Water Management Basin is its transboundary nature, as approximately 70–80% of the inflow to Lake Balkhash is provided by the Ile River, a substantial part of whose discharge is generated within the territory of China. Changes in climatic conditions in the upper reaches of the basin, together with increasing water withdrawals in the Xinjiang Uygur Autonomous Region, have a significant impact on transboundary river discharge and the water availability of Lake Balkhash. Recent studies indicate that the combined effects of climate change and increasing anthropogenic pressure in the upper Ile River basin may further reduce river inflow and exacerbate water scarcity in the downstream part of the basin [25,53,54].

The stronger correlation between the hydrological indicators and the SPEI-12 compared with shorter accumulation periods indicates the cumulative nature of hydrological drought development. This suggests that changes in river discharge result from prolonged moisture deficits rather than short-term precipitation anomalies. Similar findings have been reported by Vicente-Serrano et al. [29], Van Loon and Laaha [55], and Odongo et al. [56], who demonstrated that hydrological drought develops through the propagation of meteorological drought and is strongly influenced by the hydrological characteristics of river basins.

The response of river systems to climate change varies according to their hydrological characteristics. Small and medium-sized rivers exhibit high sensitivity to climatic variability because of the limited storage capacity of their catchments, the predominance of local sources of recharge, and the low inertia of river discharge. Under these conditions, moisture deficits and rising air temperatures are rapidly reflected in river discharge. In contrast, large rivers such as the Ile River exhibit a more regulated hydrological regime due to their extensive catchment areas and flow regulation, including the Kapshagay Reservoir. Nevertheless, even under regulated conditions, recent decades have been characterized by an increasing occurrence of hydrological droughts, indicating that the effects of climate change outweigh the mitigating influence of flow regulation. These findings are consistent with previous studies demonstrating that although river regulation reduces intra-annual discharge variability, it cannot compensate for long-term changes in water availability driven by climate change [55,57].

The integration of ground-based hydrometeorological observations with satellite remote sensing data revealed a high level of agreement between hydroclimatic and satellite-derived drought indicators. The VHI, VCI, and TCI indices reliably identified the most severe drought events detected by the SPEI, SDI, and SWSI, particularly in 2008, 2022, and 2023. These findings confirm the effectiveness of integrating remote sensing data into drought monitoring and are consistent with studies conducted in Central Asia and the Balkhash–Alakol basin, which demonstrated the high capability of the VHI, VCI, and TCI indices for assessing the spatial and temporal dynamics of drought and vegetation conditions [16,18].

This study has several limitations. The relatively small number of hydrological gauging stations may not fully capture the spatial variability of river discharge across the basin. In addition, satellite observations are only available from the early 2000s, limiting the assessment of long-term vegetation dynamics. Nevertheless, the strong agreement among the climatic, hydrological, and satellite-derived drought indicators supports the robustness of the results and demonstrates the applicability of the proposed integrated approach for drought monitoring and water resources management in the Balkhash–Alakol Water Management Basin.

5. Conclusions

The study shall present a comprehensive assessment of drought conditions based on meteorological, hydrological, and satellite data over a long-term observation period. The results obtained shall indicate a significant transformation of the region’s hydroclimatic regime, expressed in an increase in the frequency, duration, and intensity of dry periods, particularly since the late 1980s and early 1990s.

It shall be established that the rise in air temperature plays a key role in drought formation, given the absence of statistically significant changes in atmospheric precipitation. This points to the dominant influence of the thermal factor under conditions of modern climate change. It shall be shown that hydrological drought forms as a result of accumulated moisture deficit, which is confirmed by the closer correlation of hydrological indicators with indices of long-term time scales (SPEI-12). At the same time, a pronounced spatial differentiation of the hydrological response shall be identified: small and medium watercourses are characterized by high sensitivity to climatic changes, whereas large rivers, including the Ile River, demonstrate smoother runoff dynamics due to the integration of inflow and water management regulation.

Integration of satellite and ground-based data shall allow for the confirmation of the consistency in drought condition assessments and the identification that the vegetation cover response is of a more complex and indirect nature compared to hydrometeorological indicators. The results of the correlation analysis shall confirm a close interrelationship between meteorological and hydrological drought indices. At the same time, vegetation state indicators are characterized by a less direct response to changes in moisture availability and temperature.

The practical significance of this study lies in the potential to use the developed comprehensive approach to improve the drought monitoring and early warning system. The results obtained can be used to assess the risks of drought events, plan water management measures, optimize the allocation of water resources during periods of low water, and develop regional climate change adaptation programs.

The study’s findings are particularly valuable for improving water resources management in the transboundary Ile River basin, where changes in climatic conditions are compounded by increasing anthropogenic pressure and river flow regulation.

It should be noted that it is impossible to completely prevent droughts, as they are a natural manifestation of climate variability. At the same time, the timely identification of drought conditions, the development of modern monitoring and early warning systems, the improvement of water use efficiency, the optimization of reservoir operation regimes, and the implementation of adaptive management principles can significantly reduce the negative environmental, social, and economic consequences of prolonged drought periods.

Future research will focus on the integration of CMIP6 climate scenarios and advanced hydrological models to project future changes in drought characteristics within the Balkhash–Alakol Water Management Basin. Particular attention will be given to assessing the impacts of anthropogenic activities on the basin’s water resources, modeling future water availability, improving drought early warning systems, and developing scientifically based recommendations for adaptive water resources management under increasing climate uncertainty.

The results obtained will form the scientific basis for developing a regional system for monitoring and assessing drought risk, which can be used by the Republic of Kazakhstan’s state water management authorities in preparing adaptation measures aimed at enhancing the resilience of the water management system and the natural ecological systems of the Balkhash–Alakol Basin under conditions of climate change.